Two-column and sidebar CVs
Two-column CVs break most pipelines because PDF text comes out row by row across both columns: a skill from the sidebar lands in the middle of a job description, and dates end up detached from the roles they belong to. VerbatimCV groups text by position into columns, reads each column in order, and only then extracts the fields.
What goes wrong without it
- Sidebar text (skills, languages, contact details) interleaved with experience bullets.
- A date rail extracted as its own column, so
pdftotextand similar tools put all the dates after all the roles. - Section headings dropped or misplaced, so sections get misfiled.
- A language model given the scrambled text mixes up positions or invents entries to make it read sensibly.
What VerbatimCV does
- Detects columns from text positions and reads each one top to bottom.
- When the layout is ambiguous, sends an image of the page so the model can see the structure.
- Reads contact details shown only as icons by their format (email, phone, URL).
- Reads skill bars and ratings as the candidate's own opinion and flags them with the warning
SKILL_RATING_GRAPHIC. - Drops headers and footers repeated on every page.
Every experience entry keeps its own start and end dates, each with the raw
text and a precision, so you can check a date against the CV.
Try it
Get a free API key (50 parses a month, no card), then:
curl https://api.verbatimcv.com/v1/parse \
-H "Authorization: Bearer $VERBATIMCV_API_KEY" \
-F file=@cv.pdf
Agents can use the parse_resume tool instead: uvx verbatimcv-mcp. See the
quickstart for Python, JavaScript and MCP setup.